Sindhuja Thagirisa

About

LLM/agentic systems practitioner who repeatedly takes customer-facing LLM prototypes into production by operationalizing prompts, hardening RAG pipelines, and adding monitoring/guardrails. Has hands-on experience debugging intermittent production failures under high traffic (vector store timeouts/empty retrieval) and implementing fail-safe behavior plus alerting. Also partners closely with sales in pilots/POCs, customizing demos with customer data and running side-by-side comparisons to drive adoption.

Experience

Education

Key Strengths

• Turns LLM prototypes into production by decomposing systems (data sources, prompts, workflows) and hardening each component
• Strong prompt ops: externalized prompts, dynamic templating, multi-prompt strategy, and rollback/version control
• Production-grade debugging of LLM/agent pipelines using end-to-end logs and step-level instrumentation
• Root-cause analysis under real traffic conditions (staging vs production discrepancies)
• Designs safety mechanisms to prevent hallucinated/confident wrong answers when retrieval fails
• Implements monitoring/alerting (dashboards, Slack alerts, latency/timeout metrics) to catch issues early
• Customer-centric delivery: acceptance criteria, frequent check-ins, real user query mapping, parallel runs, and UAT with non-technical users
• Effective pre-sales partner: translates business needs into technical flows, customizes demos/POCs with customer data, communicates trade-offs clearly